Formatting Numbers with C++ Output Streams

David Kieras, EECS Dept., Univ. of Michigan

Revised for EECS 381, Winter 2004.
Using the output operator with C++ streams is generally easy as pie, with the only hard part being controlling the format of
the output. As you will see, this is relatively clumsy compared to C stdio's printf function, but printf is not type-safe. An
output stream has a bunch of member variables that control the details of what output looks like, and these variables can be
controlled by a set of member functions and special objects called manipulators. Most of these variables retain their values
until they are changed, but a few only act temporarily - for the next output operation only.
Throughout this document, the output stream cout is used in the examples. However, everything works for any text output
stream, such as an output le stream.
The prototypical output statement in C++ is:
cout << "Hello, world!" << endl;
This contains a manipulator, endl. This is an object, which when supplied to operator<<, causes a newline character to
be put into the output stream, followed by a call of cout's ush function, which causes the internal buffer to be immediately
emptied. This makes sure all of the output is displayed before the program goes on to the next statement. So manipulators are
objects that cause the output stream object to do something, either to its output, or to its member variables.
The manipulators with no arguments, like endl, are included in <iostream>. If you use manipulators that take arguments
(like setprecision and setw) you need to #include <iomanip>.
Like the input operator, operator<< has overloaded denitions for all of the built-in types. By default, characters and
strings are simply output as is, and this is usually satisfactory. So the main complexity is controlling the format of numbers.
There are a zillion variables that control numeric and other formats. This document focuses on how to control the most
important and common aspects of numeric output.
You can control the format by using member functions or manipulators to change a variety of member variables. This
document summarizes the most important and common things you might want to do; check a complete reference to do more.
Quite a few things can be done with the stream formatting controls, and not just with numbers; if you nd yourself tempted to
write complicated code to make the output look neat, stop and consult a good reference rst - you might be about to reinvent
the wheel.
Default output format for integers and doubles
All of the digits of an integer will be printed using decimal (base 10), with no leading zeros and a leading minus if it is
negative, using as many characters as needed, but no more.
A double value will be printed using a general format that usually works ne to represent a number in a small amount of
space. The basic amount of space used is determined by the precision. The default precision is 6, which means up to 6
signicant digits are used to represent the number. This counts digits both to the left and the right of the decimal point. Fewer
digits are printed if the result is accurate; e.g. trailing zeros at the right of the decimal point are not printed. As shown in the
example below, as the number increases in size, places to the right of the decimal point will be dropped, and the result
rounded off as needed to stay within 6 digits. If the number cannot be represented with six digits to the left of the decimal
point after all to the right have been dropped, the output will ip into exponential (scientic) notation showing 6 signicant
1
digits and the exponent of 10. You can increase or decrease the precision, and this will change the number of signicant digits
shown in the output.
The problem with the general format is that if you want to make numbers line up in neat columns, as in a table of data, the
general format works very poorly because the total number of characters, and the number of digits to the right of the decimal
point, will vary depending on the size of the number. The rest of this handout descri bes the basic techniques for controlling
the format to produce neat-looking output.
Saving and restoring stream settings
First, if you are going to change the format settings of the stream, you normally will want to restore them to what they were
before you changed them, so that different parts of your program will not interfere with each other. You can get the current
settings (called "ags") from the output stream with a member function named flags. The settings can be stored in a type of
object called a fmtflags object, dened in a class called ios, which is included with <iostream>. You can declare one of
these objects, but you have to declare it using the scope resolution operator. To make a long story short, the following
statement will save certain aspects of the format state in the variable old_settings:
ios::fmtflags old_settings = cout.flags();
Ugly! Then, after doing the output using the new setting, you can restore the old setting by calling the same function with the
old settings as an argument:
cout.flags(old_settings);
Other settings can be obtained and restored with member functions. For example,
int old_precision = cout.precision();
will save the current precision specication. Then
cout.precision(old_precision);
will restore the precision to the original value.
Controlling minimum eld width
You can control the minimum number of characters used to print the number by specifying the eld width. Leading blanks
will be included to ll up the eld, but if the number doesn't t, the extra digits will be printed regardless of the eld width.
The eld width also affects output of strings.
Important: Field width changes are only temporary; they affect only the next relevant output. Changing the width affects
only the immediately following output of a number, whereupon the width setting automatically reverts to the standard
behavior of "as many characters as needed" (specied by a eld width of zero).
For example:
cout << "*" << setw(4) << 12 << "*" << endl; // manipulator
will produce:
* 12*
There are two spaces between the '*' and the '1' for a total of 4 characters between the '*'s.
2
We can do the same thing with a member function, but we have to call the member function right before the relevant item is
output. If we called it before outputting the rst "*", the new width would control the output of the "*", and then would reset,
and so not affect the output of the number.
So the following does the same thing as the previous example:
cout << "*";
cout.width(4); // member function
cout << 12 << "*" << endl;
Precision and the general oating-point format
You can change the maximum number of signicant digits used to express a oating point number by using the precision
member function or manipulator. For example,
cout.precision(4); // member function
cout << 1234.56789 << " " << 245.678 << " " << 12345.0 << endl;
or
cout << setprecision(4) // manipulator
<< 1234.56789 << " " << 245.678 << " " << 12345.0 << endl;
will produce:
1235 245.7 1.234e+04
Notice that when places to the right of the decimal point are dropped, the result is rounded off. If the number is too large to be
represented in normal notation in that precision, as in the last value, the precision specication is ignored and the value is
output in a more general form. The precision stays changed until you change it again.
Setting the precision to zero with:
cout.precision(0);
or
cout << setprecision(0);
restores the 6-digit default.
Precision and the xed oating-point format
For neat output of doubles, the xed format is most useful. (You can also select a scientic notation format.) In xed format,
the precision species the number of digits to the right of the decimal point, and a precision of zero means zero places to the
right of the decimal point (i.e. round to the nearest unit).
Using xed format together with setting the minimum eld width allows one to arrange numbers that take up a uniform
number of characters, providing that they all t in the minimum eld width.
Using a member function to set xed oating-point format is incredibly ugly. The setf member function is used to set a
certain bit in the format ags, using constants dened in ios:
3
cout.setf(ios::fixed, ios::floatfield);
Fortunately, you can select the xed format with a simple manipulator:
cout << fixed;
You can reset the oating-point format ags to the default with
cout.setf(0, ios::floatfield);
but usually you will want to restore the previous settings.
An Example of Controlling Numerical Output Format
In the following examples, we compute the valaue of pi raised to various negative and positive powers, and output the
exponent and the result. The example shows the effects of setting precision, then what we get when we set the xed format
and then use different precisions and output widths. By using a big enough eld width to cover the range of values with a
xed format, we can get a neat tabular output result. The example saves and restores the default settings as an example. Each
batch of output has been pasted into the code as a comment with some discussion to make the example easier to follow. This
code has also been posted in the examples section of the course web site.
#include <iostream>
#include <iomanip> // needed to use manipulators with parameters (precision, width)
#include <cmath> // needed for pow function
using namespace std;
int main ()
{
const int beginvalues = -10;
const int endvalues = 16;
const int nvalues = endvalues - beginvalues;

int ipow[nvalues];
double ary[nvalues]; // an array for demo values

int ipowindex = 0;
// fill array with interesting range of values
for (int i = beginvalues; i < endvalues; i++) {
ipow[ipowindex] = i;
ary[ipowindex] = pow(3.14159265, i);
ipowindex++;
}

// output index and array[index] using default settings
cout << "Output using default settings" << endl;
for (int i = 0; i < nvalues; i++)
cout << ipow[i] << ' ' << ary[i] << endl;
/*
Output using default settings
-10 1.06783e-05
-9 3.35468e-05
-8 0.00010539
-7 0.000331094
-6 0.00104016
-5 0.00326776
-4 0.010266
4
-3 0.0322515
-2 0.101321
-1 0.31831
0 1
1 3.14159
2 9.8696
3 31.0063
4 97.4091
5 306.02
6 961.389
7 3020.29
8 9488.53
9 29809.1
10 93648
11 294204
12 924269
13 2.90368e+06
14 9.12217e+06
15 2.86581e+07
Each output double value has either six significant digits or fewer if the value can be
expressed just as accurately. For example, for ipow = 0, the value of one shows with no
decimal places and not even a decimal point. These are not printed in the default format
unless there are non-zero places printed to the right of the decimal
point. See also ipow = 10 through 12, where the decimal places have all been rounded off.
At i = 13 and beyond, 6 digits are not enough, so the output flips into scientific notation,
still showing six significant digits, but with an exponent of ten. A different rule, not
so easy to state, governs when small values flip into scientific notation.
*/
// save the current settings
ios::fmtflags old_settings = cout.flags(); //save previous format flags
int old_precision = cout.precision(); // save previous precision setting
// don't need to save width setting, because it automatically resets to default value
// after each numeric output item.

// just change the precision
cout << setprecision(8);
cout << "\nOutput using integers with default output" << endl;
cout << "doubles in general format, precision 8" << endl;
for (int i = 0; i < nvalues; i++)
cout << ipow[i] << ' ' << ary[i] << endl;

/*
Output using integers with default output
doubles in general format, precision 8
-10 1.0678279e-05
-9 3.3546804e-05
-8 0.00010539039
-7 0.00033109368
-6 0.0010401615
-5 0.0032677637
-4 0.010265982
-3 0.032251535
-2 0.10132118
-1 0.31830989
0 1
1 3.1415927
2 9.8696044
3 31.006277
5
4 97.409091
5 306.01968
6 961.38919
7 3020.2932
8 9488.5309
9 29809.099
10 93648.046
11 294204.01
12 924269.17
13 2903677.2
14 9122171
15 28658145
Here up to 8 significant digits are printed, which is enough to avoid the scientific notation
at ipow = 15. The result is still a mess because the values take up different numbers of
spaces.
*/
// change output format settings with member functions
cout.setf(ios::fixed, ios::floatfield); // set fixed floating format
cout.precision(2); // for fixed format, two decimal places
// cout << fixed << setprecision(2); // same effects, but using manipulators
cout << "\nOutput using integers with width 2," << endl;
cout << "doubles in fixed format, precision 2, width 8" << endl;
for (int i = 0; i < nvalues; i++)
cout << setw(2) << ipow[i] << ' ' << setw(8) << ary[i] << endl;
/*
Output using integers with width 2,
doubles in fixed format, precision 2, width 8
-10 0.00
-9 0.00
-8 0.00
-7 0.00
-6 0.00
-5 0.00
-4 0.01
-3 0.03
-2 0.10
-1 0.32
0 1.00
1 3.14
2 9.87
3 31.01
4 97.41
5 306.02
6 961.39
7 3020.29
8 9488.53
9 29809.10
10 93648.05
11 294204.01
12 924269.17
13 2903677.23
14 9122171.04
15 28658145.48
All the double values show two places to the right of the decimal point, with the results
rounded to hundredths (possibly to zero). This output would be quite neat except for two
problems: (1) Since the minus sign counts in the width of the integers, the -10 value won't
fit into two spaces, and this messes up the first line. (2) Starting at ipow = 11, the output
6
is messed up because the results will not fit into the total space of 8 characters
(the decimal point counts as one space). Note that setting fixed format prevents the flipping
into scientific notation, and forces the value of exactly one to show with a decimal point
and the specified number of places to the right of the decimal point.
*/
cout << "\nOutput using integers with width 3 integers, " << endl;
cout << "doubles in fixed format, precision 0, width 5" << endl;
// can use manipulators to change precision and others inside an output statement
// in fixed format, precision of zero means no decimal places
for (int i = 0; i < nvalues; i++)
cout << setw(3) << ipow[i] << ' ' << setprecision(0) << setw(5) << ary[i] << endl;
/*
Output using integers with width 3 integers,
doubles in fixed format, precision 0, width 5
-10 0
-9 0
-8 0
-7 0
-6 0
-5 0
-4 0
-3 0
-2 0
-1 0
0 1
1 3
2 10
3 31
4 97
5 306
6 961
7 3020
8 9489
9 29809
10 93648
11 294204
12 924269
13 2903677
14 9122171
15 28658145
This gives room for the negative integer values, and so it produces a neat output until
ipow = 11, whereupon the output takes additional digits just as in the previous example.
Because the fixed precision is zero, everything is rounded to the nearest integer value,
and thus neither a decimal point nor places to the right of the decimal point appear.
For values less the one, of course, the result rounds off to zero.
*/
cout << "\nOutput using integers with width 3 integers, " << endl;
cout << "doubles in fixed format, precision 8, width 18" << endl;
cout << setprecision(8);
for (int i = 0; i < nvalues; i++)
cout << setw(3) << ipow[i] << ' ' << setw(18) << ary[i] << endl;

/*
Output using integers with width 3 integers,
doubles in fixed format, precision 8, width 18
-10 0.00001068
-9 0.00003355
7
-8 0.00010539
-7 0.00033109
-6 0.00104016
-5 0.00326776
-4 0.01026598
-3 0.03225153
-2 0.10132118
-1 0.31830989
0 1.00000000
1 3.14159265
2 9.86960438
3 31.00627657
4 97.40909059
5 306.01968304
6 961.38918698
7 3020.29320362
8 9488.53092933
9 29809.09902689
10 93648.04640600
11 294204.01427594
12 924269.16884980
13 2903677.22748013
14 9122171.03582395
15 28658145.47818740

This output is the first that is completely neat over the whole range of values.
The width for the doubles leaves enough room for the additional digits to the left
of the decimal point. Of course, the precision of 8 produces a lots of decimal places
which we may not need.
*/

// restore output format flags and precision
cout.flags(old_settings);
cout.precision(old_precision);
cout << "\nOutput using original settings" << endl;
for (int i = 0; i < nvalues; i++)
cout << ipow[i] << ' ' << ary[i] << endl;
/*
Output using original settings
-10 1.06783e-05
-9 3.35468e-05
-8 0.00010539
-7 0.000331094
-6 0.00104016
-5 0.00326776
-4 0.010266
-3 0.0322515
-2 0.101321
-1 0.31831
0 1
1 3.14159
2 9.8696
3 31.0063
4 97.4091
5 306.02
6 961.389
7 3020.29
8 9488.53
9 29809.1
8
10 93648
11 294204
12 924269
13 2.90368e+06
14 9.12217e+06
15 2.86581e+07
*/
}
9